Abstract 16374: The Vancouver General Hospital Pericardial Effusion Risk Index

• Published in: Circulation (New York, N.Y.) N.Y.), 2016-11, Vol.134 (Suppl_1 Suppl 1), p.A16374-A16374
• Main Authors: Tong, Calvin Ka Wing, Gin, Ken, Jue, John, Lee, Pui Kee, Nair, Parvathy, Szefer, Elena, Thompson, Darby, Tsang, Michael, Tsang, Teresa
• Format: Article
• Language: English
• ISSN: 0009-7322; 1524-4539

BackgroundManagement of hemodynamically stable patients with at least moderate sized pericardial effusion is often at the discretion of attending cardiologists. Objective scoring index to guide pericardiocentesis was previously developed in a small population of 48 patients in New York. We aim to harness the large patient volume at the Vancouver General Hospital (VGH) to develop our own pericardial effusion risk index (PERI) applicable to a large Canadian tertiary care centres.MethodsA conditional inference random forest was used to predict which patients received pericardiocentesis. Candidate variables for the classifier included patients’ demographics (age, sex), clinical characteristics (heart rate, blood pressure), presenting symptomology, comorbidities, pericardial effusion etiology (cancer, trauma, inflammatory, idiopathic, etc), hemodynamic variables on echo studies (RA collapse, RV collapse, mitral inflow velocities, IVC plethora), and size of effusion. Performance of the random forest was assessed using an optimism-adjusted AUC based on 200 bootstrap samples; which should represent the AUC in an independent validation set. Importance of variables is based on the mean decrease in classification accuracy if that variable was removed.ResultsA total of 140 patients (48% women) were included. Of these, 58 underwent echo-guided pericardiocentesis. Patients’ age ranged from 17 to 96. The size of effusions ranged from 10 to 43 mm (mean 20 mm) with 32% of subjects experiencing RA collapse > 1/3 of the cardiac cycle. A total of 116 subjects with no missing data for 32 of the variables of interest were used to determine the risk index. The VGH PERI had excellent ability to discriminate those who received pericardiocentesis from those who did not, with an optimism-adjusted AUC equals to 95.9. The most important predictors were the size of effusion, RV diastolic collapse, mitral inflow velocity at the first cardiac cycle after the beginning of inspiration, and dyspnea.ConclusionThe VGH PERI allows prediction of the need for pericardiocentesis in hemodynamically stable patients with at least moderate sized effusion. The index has been internally validated in a large sample size, and will be undergoing external validation.